• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Next-Gen PS5 & XSX |OT| Console tEch threaD

Status
Not open for further replies.
Obviously they hold very different positions at their perspective companies but where their rolls converge is being kinda the spokesperson for their platform.

Phil’s input on the hardware was from a financial and strategic standpoint. Cernys input on the hardware is getting the best out of it given the financial and strategic guidelines passed down to him.

When either of them get in front of a camera it’s to be a “salesman”

That's fair but Cerny is actually involved in the design of the console so he's "selling" something he had a direct hand in developing, Spencer isn't going to be that hands on, that'd be like saying Jim Ryan helped design the PS5.

I'd also say that Spencer is more of a spokesman for the Xbox brand, he speaks at all of their press conferences etc, Cerny mostly only speaks when it's directly related to the hardware design etc otherwise it's someone else from Sony who does all of the talking. We never saw Cerny go on stage at E3 etc but Spencer always did, even before he was promoted to the job he has now.
 

FeiRR

Banned
I've made a quick comparison of HDD vs SDD loading times on a PS4 Pro with a cheap external SSD drive. I'm using ADATA External SD700 512G drive which cost me less than 100 EUR, connected to PS4 Pro via the back external USB port. The test was done with Division 2 installed on HDD and then moved to SSD. It's the same game, of course it's not patched in any way to benefit from SSD. What I'm doing in that video is:
- loading the game from the main menu screen to gameplay (I skipped the logo animations at the beginning because they play the same speed on both so no reason)
- fast traveling from DC location to NYC location (game's DLC) and back
I didn't include any in-game loadings because during missions the game hides loading with doors, lifts and other action-stoppers. They don't play any faster so this would show nothing.

The results were as follows:
  • HDD results:
    • title screen to DC: 148 s
    • DC to NYC: 60 s
    • NYC to DC: 96 s
  • SSD results:
    • title screen to DC: 76 s
    • DC to NYC: 24 s
    • NYC to DC: 38 s
As you can see, we can observe 2 to 3 times improvement in loading. Remember that SSD over USB in PS4 Pro is a quick and cheap band-aid solution, which doesn't even utilize the full speed of that SSD drive (around 250 MB/s benchmarked on PC) due to data bus bottlenecks. Now imagine speeds of 10, 20, 50 times more.

The video is here.
 

ZywyPL

Banned
(You don't need to do any extra work as a dev, the system (API, I/O, etc) is programmed to load only assets needed within your sight like on Horizon Zero Dawn but on steroids (more room for CPU/GPU optimization for more graphics and computations with less to no waste).

They will have to, a lot of it actually. Crytec said some time ago after initial spec reveals that because of SSDs most games engines will have to be heavily modified and the devs will have to learn a lot in order to fully utilize the tech for something other than faster loadings or less pop-up.

That's part of the solution for slower systems. In PS5 you don't even need load things that you MUST load on other systems, resulting in much more headroom for the CPU/GPU/RAM to do more where needed.

Spreading FUD an misinformation against XBX yet again I see, those two bans didn't learn you anything huh? That fine-grained polygon rendering is what Geometry Engine is responsible for, which both consoles are equipped with.


Sosokrates Sosokrates he's been spreading the same FUD in another thread.

Just report him so mods can take notice, he's clearly not letting it go. It's only a matter of time when he will try to convince people that the two additional SSD I/O co-processors are also part of the 36CUs.


Yeah, and how is that 2.4GB/s doing? That 10GB RAM will get really crowded. And no, XSX can't EVEN dream of what PS5 can do, so far it's barely 4.6x faster than HDD with hard evidence provided by Microsoft themselves. Ever heard of bottlenecks?

And how is PS5 doing with 10x faster loading with 90x faster than PS4 drive? Besides, I thought it wasn't about the loading times at all but, what's it called? A new paradigm in design, a real game-changer? You spin fasted than a fidget spinner on a drill, stick with one narrative please.

You can nitpick titles as much as you want. It doesn't change the fact that they aren't just going to wait for everyone to transition,.

The publishers won't lose billions of dollars each year because some random on NeoGAF says so, LMFAO. See below for more explanation.

The changes in level design for the use of SSD will not works in pc without a SSD or some another
solution like cache a huge quantity of data in RAM.

Really how old are you ? Is the first change of generation you live of what ? Because this sound to me like a person only who only listen a youtuber which build
pc as business or something similar.

You just basically confirmed with what he says and what you're arguing about the whole time - because of HDD in PCs there won't be much/any changes in the level designs, I said it already weeks ago that whoever believes in absolutely NO loading screens lives in alternate reality, loading screens are here to stay, the difference is that PS5 will have them displayed for for example 3-5s, XBX for 5-7s, and PC anything from half a second to 2min, depending on the specs. Loading screens is what will ensure compatibility with all devices out there, just a single bitmap will allow the publishers to reach entire customer base, no one is leaving all those 150MLN PS4/XB1 users behind, let alone the PC crowd. All those multi-billion dollar companies like EA, Ubi, Acti etc. are multi-billion dollar companies in the first place exactly because they don't follow biased fanboy agendas. The PS5 SSD won't be utilized at its fullest aside PS5 1st party exclusives, but expecting the same from 3rd party publishers is simply bending the reality. I think half a generation with nothing but cross-gens and indie ports still wasn't enough for some to open their eyes an realize how this whole business works and what it is about (spoiler: it's money), I mean, the "PS4 has no games" meme is just 3-4 years old...

I think Bo_Hazem is trolling again

Yeah, it took him what, 2-3 days after the unban and he's at it once again, I already reported him, some people just never learn no matter how many chances you give them.

It's like Acend didn't watch Cerny's presentation at all. Nobody seems to have.

The SSD was designed to address those issues and limitations. So you get the exact speed. No bottlenecks and other usual limitations. Stop comparing brute PC hardware to custom built consoles.

Everyone watched it, but no one saw any actual games, or even tech-demos. So until proven, Cerny's words are just a marketing/PR empty BS/buzzwords. How did it go? "Don't talk, show"?

PS5 I/O has fine grained modifications to minimize the impact of random reads and Cerny gave a typical in game performance metric of 8-9GB/s which accounts for random reads in gaming.

Then why did he compared it with HDD's sequential speeds? Doesn't make any sense...
 
Last edited:
I've seen a lot of concern that games especially multiplats won't be using NVMe Speed to full effect.
I beg do differ, I think developers will now have incentive to take this into consideration.
So they'll probably have all those different speed types hdd/sdd/nvme in mind when designing and creaing their engines/games.
Just like you have options nowadays for different GPU and CPU usage in your settings, those settings might actually even depend on the harddrives thats beeing used.

You have a hdd and [...]? Can't use the graphic options x,y,z -> Play on potatoe level
You have a ssd and [...]? You can use the graphic option x but not y,z -> Play on lower/mid graphic level
You have a nvme and [...]? You can use the gprahic option x,y but not z -> Play on mid/high graphic level
You have a high end ocie4 nvme and high end graphics card with RT and AI cores and such shit? Your can haz all the grphics -> Play on Ultra graphic level
 
Last edited:

Sosokrates

Report me if I continue to console war
What exactly are you trying to achieve here? That it's 10tf and not 10.3tf? Is that 0.3tf such a big deal to you to keep going on and on about it?

Keep going on about?
Im just replying to drive by trolls and ad hominem attaacks.
If people were cival and actually discussed with resorting to the above we would not have this issue.
It strange that you assume negativity.
I am generally interested what the tempest engine is, but people are so insecure that unless its 100% positive ps5 talk they just attack it is a real shame.
 

Shmunter

Member
I've seen a lot of concern that games especially multiplats won't be using NVMe Speed to full effect.
I beg do differ, I think developers will now have incentive to take this into consideration.
So they'll probably have all those different speed types hdd/sdd/nvme in mind when designing and creaing their engines/games.
Just like you have options nowadays for different GPU and CPU usage in your settings, those settings might actually even depend on the harddrives thats beeing used.

You have a hdd and [...]? Can't use the graphic options x,y,z -> Play on potatoe level
You have a ssd and [...]? You can use the graphic option x but not y,z -> Play on lower/mid graphic level
You have a nvme and [...]? You can use the gprahic option x,y but not z -> Play on mid/high graphic level
You have a high end ocie4 nvme and high end graphics card with RT and AI cores and such shit? Your can haz all the grphics -> Play on Ultra graphic level
I think quick adoption of ssd will come from ease of use. It will be easier to code and access those assets rather than bumping up against traditional constraints, needing additional problem solving and workarounds.

Faster development, better results, cheaper dev costs will be the catalyst.
 
Last edited:

Sosokrates

Report me if I continue to console war
So explain then your thought process, how does your 35CU + TE theory alings with known info. For starters compute performance

Ive already explained like over 5 times. Its really getting unenjoyable, several posters quoting me with same stuff.

But its mainly based on what df have said.

a0211f1340898826.jpg


Here they even say "revamped GPU Compute unit" and its the same clockspeed as the gpu.

While this does not my proove my theory it does support it.

What is quite sad about this place is that i see to be the only one providing sources for my arguments, wherae as most of the people disagreeing with me dont, instead they just come up with ad hominems, or just try to attack my character.



@18.30
 

THE:MILKMAN

Member
I think quick adoption of ssd will come from ease of use. It will be easier to code and access those assets rather than bumping up against traditional constraints, needing additional problem solving and workarounds.

Faster development, better results, cheaper dev costs will be the catalyst.

The bold might just be the key here if I (as a layman) understand what Mark Cerny says below. It sounds too good to be true/a dream for devs but the whole I/O system blocks seem to be designed to allow the below to be done automatically!?

Mark Cerny said:
The best thing is as a game developer when you read from the SSD you don't need to know any of this. You don't even need to know if your data is compressed. You just indicate what data you would like to read from your original uncompressed file and where you'd like to put it and the whole process of loading it happens invisibly to you and at very high speed.
 

mitchman

Gold Member
"DirectStorage is an all new I/O system designed specifically for gaming to unleash the full performance of the SSD and hardware decompression. It is one of the components that comprise the Xbox Velocity Architecture. Modern games perform asset streaming in the background to continuously load the next parts of the world while you play, and DirectStorage can reduce the CPU overhead for these I/O operations from multiple cores to taking just a small fraction of a single core; thereby freeing considerable CPU power for the game to spend on areas like better physics or more NPCs in a scene. "

Sound at all like PS5's solution? What about this?
Doesn't sound at all like the PS5 solution. No DMA and GPU cache scrubbers is mentioned, which is a corner stone of the performance.
 

Darius87

Member
Ive already explained like over 5 times. Its really getting unenjoyable, several posters quoting me with same stuff.

But its mainly based on what df have said.

a0211f1340898826.jpg


Here they even say "revamped GPU Compute unit" and its the same clockspeed as the gpu.

While this does not my proove my theory it does support it.
what you're saying is desperate hater and attention seeking fanboy nonsense there's 0% chance that it does support what you're saying
  1. RDNA.2 doesn't support odd number of CU's because of its double CU config.
  2. Neither DF or Sony have claimed that so i don't know why are you even refering to DF video in first place?
  3. audio chip doesn't have cache and it 's redesigned for audio purposes how it would be be on the same die with apu is beyond stupid.
  4. if sony would wanted audio on it's apu it would be runing on cpu ar GPGPU(compute).

What is quite sad about this place is that i see to be the only one providing sources for my arguments, wherae as most of the people disagreeing with me dont, instead they just come up with ad hominems, or just try to attack my character.
because it's nonsense not sources i could provide you sources and say xsex is actually 10 Tflops not 12. it doesn;t make any sense.
 
Last edited:

ZywyPL

Banned
I'll ask you the same question then. Publishers had no issue leaving those 150+ million PS3/360 users behind. Why do you honestly believe that's not going to be the case again this gen?

In which universe/reality? For the first few years PS4/XB1 were getting mostly cross-gens/remasters from PS3/X360, which there was a ton of bitching back then about how they are holding back the new hardware. So I revert the question - why do you honestly believe that's not going to be the case again this gen?
 
In which universe/reality? For the first few years PS4/XB1 were getting mostly cross-gens/remasters from PS3/X360, which there was a ton of bitching back then about how they are holding back the new hardware. So I revert the question - why do you honestly believe that's not going to be the case again this gen?

Wrong. After its first year. The big pubs were pretty much making current gen exclusives.
 

Shmunter

Member
Ive already explained like over 5 times. Its really getting unenjoyable, several posters quoting me with same stuff.

But its mainly based on what df have said.

a0211f1340898826.jpg


Here they even say "revamped GPU Compute unit" and its the same clockspeed as the gpu.

While this does not my proove my theory it does support it.

What is quite sad about this place is that i see to be the only one providing sources for my arguments, wherae as most of the people disagreeing with me dont, instead they just come up with ad hominems, or just try to attack my character.



@18.30


Wow, the trolling needs to stop. Cerney in his address refers to Tempest as a ‘hardware unit’ they’ve built. Even discusses how it is DMA enabled etc. A hardware unit = component by any defining measure.

The terminology, the tech discussed, the basic principles of common sense, disqualifies this being a reserved GPU CU.

Just because the PS5 offers certain custom hardware over your preferred console won’t end your world, even if it may feel like it to you. Go cry in the corner troll, and stop leaving your shit stain all over the place. Pages and pages of the same wasted space.
 
Last edited:

Ascend

Member
You're moving the goalpost after you found out that you were wrong.



See that?

You said games do not use more than 6GB of VRAM today.

You weren't talking about 6GB of VRAM being enough, you said 4K games do not use MORE than 6GB of VRAM.


As shown on Resident Evil 2 Benchmarks, it exceeds that limit. 6GB.



Want another game?



Doesn't matter if the VRAM usage wasn't 100% accurate, it still requires more than 6 if you want to best visuals possible.

Just let me rephrase to reflect better what I mean... Today, games at 4K do not need more than 6GB to run at max settings. There. As for the rest of your kerfuffle;

giphy.gif


Stop the negativity guys, discuss things without fighting.
Respect both sides of the argument and explain your facts/theories/ideas.
Kind of hard to do when people are deliberately putting words in your mouth and twisting your words to pretend to win an argument.

Doesn't sound at all like the PS5 solution. No DMA and GPU cache scrubbers is mentioned, which is a corner stone of the performance.
Do you even know what those things are? Even the Xbox One has DMA. You really think the XSX will not?
As for cache scrubbing... I'll leave this here, and let people themselves draw conclusion on what the influence is on performance... And whether it really is some fancy PS5 only tech...

 
Last edited:

DaGwaphics

Member
Yes both are huge, but only the PS5 offers speeds yet unseen. The PS5 ssd is essentially 800 terrabytes of DDR3.

Say what now? A typical dual channel DDR3 setup will net you 26GB/s, and that's a cheap and slow setup. Jack up the clocks and you can get 60GB/s+ from DDR3. 5.5GB/s is barely on par with your average DDR setup from 2000. Then there's the little issue of not having bit level access, and the fact that max speed will be for sequential access only. The SSD isn't acting as RAM in the traditional sense for these systems, it's simply too slow. 800TB sounds like a great drive though, plenty of room on that sucker. I hope they can keep the price under 500k.
 
You are completely missing the point that an entry-level 512 GB SSD is about $50, so less than one RRP game. It's also a very easy to install and start benefiting from its performance: you just install a game to it and there you go. I switched to SSDs in my workstation years ago and never looked back. Those were simple, inexpensive upgrades which significantly increased the comfort of work. Next gen titles will just have an SSD as a requirement. There's even a PC exclusive game which requires SSD right now.

Steam hardware stats page says you're very wrong. It's not a good way to start a membership with groundless conjecture and a passive-agressive attitude towards your interlocutors. Just a friendly advice, we all have a 'report' button and as a new account, you're on probation.


KZ:S reflections aren't RT. There are several ways to make them look like that in realtime without RT.

You can use cube mapping, which is throwing a static texture in the mirror which resembles what would be reflected in it. It's very easy and doesn't take any additional resources when your game is running but in game production you have to render those textures first. Infamous Second Son used that approach: window reflections were calculated on fast render farms to be put into the game. The effect is great but the preparation phase required a lot of artists working on that, which translates into hefty budget.

Another technique used in many games to make puddles and other water volumes is screen space reflections. It can render dynamic surroundings of a said reflective surface in realtime, even including the character models. It's the closest to RT in terms of quality but also has several limitations, the biggest of them probably being that it can only reflect what you see on screen (you as a player), so won't work with mirrors that you look straight at because then you'd see the cameraman... No, I'm joking ;) You wouldn't see anything behind virtual camera of the game. It's also very taxing on hardware. That's why a lot of mirrors or puddles in games aren't very detailed (render at lower resolution and framerate) while a lot of mirrors are broken. This is to hide the lesser quality of that image. If you remember the first SpiderMan trailer, it has a lot of puddles, the construction yard was all wet. They decided to remove most of the puddles because it was too taxing on hardware and didn't add that much to the whole scene.

Finally, you can just make all objects behind the mirror/window real, just put them there, add a shader to simulate glass and there you go. Of course it's also very taxing on hardware but works well in closed spaces, like baths where a character looks into their reflection. Then you have two character models and a hole in the wall. This technique is probably the oldest. The first time I remember seeing a reflection in a 3D game was Duke Nukem. At that time it even wasn't a 3D model, just a sprite.


That's an interesting take. With all the effort to speed-up I/O, there must be a real problem with encryption. PS4 encrypts/decrypts data on-the-fly. It's quite easy to do if you have a 50MB/s data stream. IIRC, one Jaguar core is dedicated to that (in PS3 it was one SPU). I don't think even a Ryzen CPU core can do on-the-fly encryption of 5 GB/s. Or can it?

I think it did, to a degree. But only for certain glossy reflections though. Eurogamer did an interview with one of the games' directors and he described he they had gone about doing it. The article can be found here, it's a very interesting read.
 

Shmunter

Member
Say what now? A typical dual channel DDR3 setup will net you 26GB/s, and that's a cheap and slow setup. Jack up the clocks and you can get 60GB/s+ from DDR3. 5.5GB/s is barely on par with your average DDR setup from 2000. Then there's the little issue of not having bit level access, and the fact that max speed will be for sequential access only. The SSD isn't acting as RAM in the traditional sense for these systems, it's simply too slow. 800TB sounds like a great drive though, plenty of room on that sucker. I hope they can keep the price under 500k.
Are you sure you’re not referring to DDR4? Also sequential access? This is an SSD with 0 seek times, not a platter HDD.

Source... https://www.crucial.com/support/memory-speeds-compatability

DDR3 Specs
DDR3-800 PC3-6400 - 6400 MB/s
DDR3-1066 PC3-8500 - 8533 MB/s
DDR3-1333 PC3-10600 - 10667 MB/s
DDR3-1600 PC3-12800 - 12800 MB/s
 
Last edited:

DaGwaphics

Member
Are you sure you’re not referring to DDR4? Source... https://www.crucial.com/support/memory-speeds-compatability

DDR3 Specs
DDR3-800 PC3-6400 - 6400 MB/s
DDR3-1066 PC3-8500 - 8533 MB/s
DDR3-1333 PC3-10600 - 10667 MB/s
DDR3-1600 PC3-12800 - 12800 MB/s

No, those are the single channel metrics. Only the cheapest embedded boards ever used that.

Edit: Here's the first old chip I could find in Google, notice the dual channel max DDR3 1600 bandwidth listed there. https://ark.intel.com/content/www/u...core-i3-4130-processor-3m-cache-3-40-ghz.html

Also there are still extreme differences in both iops and bandwidth on SSDs in terms of sequential R/WR vs. random R/WR. Also, don't look at the maximum ops on DDR3 and compare that to what is available on SSD, the results will frighten you.
 
Last edited:

TLZ

Banned
Keep going on about?
Im just replying to drive by trolls and ad hominem attaacks.
If people were cival and actually discussed with resorting to the above we would not have this issue.
It strange that you assume negativity.
I am generally interested what the tempest engine is, but people are so insecure that unless its 100% positive ps5 talk they just attack it is a real shame.
I'm sorry but I don't believe you one bit. The exact numbers were given to you by Cerny, yet you still want to create an argument that doesn't exist. They told you gfx will use 36 CUs at 2.23ghz giving you 10.28TF. But you desperately want to prove it's 35 CUs and 10TF. Why?

Wasn't there an article or something similar (I can't remember exactly what it was) that discussed this extra repurposed CU that acted like an SPU that used so much power, that if normal CUs were used instead, it'd take up 7 CUs alone? (I think the number was 7 I can't remember).
 

Shmunter

Member
No, those are the single channel metrics. Only the cheapest embedded boards ever used that.

Edit: Here's the first old chip I could find in Google, notice the dual channel max DDR3 1600 bandwidth listed there. https://ark.intel.com/content/www/u...core-i3-4130-processor-3m-cache-3-40-ghz.html
Yes, dual channel is legit. But the PS5 ssd is still undoubtedly in ballpark off a single channel ddr3 setup. Of which many pc out there also did at the time, both are true at the same time.

But even at it’s single channel ddr3 speed, it surely must be acknowledged it can certainly offer plenty fast assets to feed the GPU In streaming scenarios, and a massive amount of it to boot.

Edit : I see you’ve added a bit about ssd sequential vs random access speed differences. I’ve never heard of this applying to ssd. It certainly doesn’t make sense without ssd having mechanical restrictions, but I’m no expert.
 
Last edited:
Say what now? A typical dual channel DDR3 setup will net you 26GB/s, and that's a cheap and slow setup. Jack up the clocks and you can get 60GB/s+ from DDR3. 5.5GB/s is barely on par with your average DDR setup from 2000. Then there's the little issue of not having bit level access, and the fact that max speed will be for sequential access only. The SSD isn't acting as RAM in the traditional sense for these systems, it's simply too slow. 800TB sounds like a great drive though, plenty of room on that sucker. I hope they can keep the price under 500k.

Theoretical peak

They don't get anywhere close to those numbers in practicality
 
Last edited:

DrKeo

Member
The die of the PS5's APU actually has a total of 40 CUs in the portion that is the GPU. However, due to the yield rate of the APU causing all 40 CUs to be functional on only very few units of the APU, the APU is designed to utilize less than the total number of CUs. Hence, it's possible that the number of CUs that it is designed to use is 37: 36 for rendering tasks and 1 for audio computations.

However, I'm sure that I've read that the Tempest Engine is an independent hardware component (i.e. a chip) that's separate from the APU or at least separate from the CPU and the GPU but still within the APU. Whichever may be the case, the Tempest Engine is not one of the 36 CUs that are dedicated to graphics tasks. As for why it runs at the same frequency of the GPU, perhaps it has to do so in order to sync the audio with what's being rendered.
We dont have confirmation, so it could be one of the 36cus, it is unlikely that there will be only 3 unactive cus. Sony wont want worse yields.
And it’s unlikely it’ll be a stripped non graphics CU within the same 36 block.

Considering we have the TF amount that’s not based on 35, but 36, no?
So even though every site says PS5 has 36 CU active, you're thinking if the system has some disabled CUs in good condition, some PS5 systems might have 37 CU to work with?

So some systems will work better than others?
Oh the irony

lol come on you are just trolling at this point, how do propose they reach 10.27TF with 35CUs, another one of Cerny's lies? 🤥
TE uses a stripped down CU as its frame it does not contribute to the compute throughput, its independent of the 40CUs.
You are like 3rd person to just drive by and be like "oh just trolling"

These petty accusations achieve nothing.

And your quotes of me make no sense.

You are literally just repeating questions others have said, try reading the thread. And dont waste my time
So explain then your thought process, how does your 35CU + TE theory alings with known info. For starters compute performance
This sounds like bollocks. There's no way that Sony would design the PlayStation 5 to have less CUs dedicated to graphics rendering than the Playstation 4 Pro; I believe that they actually designed the PlayStation 5 to use 36 CUs for graphics rendering partly because it would run PlayStation 4 games more easily than it would otherwise, since 36 CUs is perfectly divisible by the number of functional CUs in the base PlayStation 4 (18CUs, which is exactly half of 36) and is the same number of CUs that the PlayStation 4 Pro uses for graphics rendering.

So, the argument that the PlayStation 5 uses only 35 CUs for graphics rendering is ridiculous and is an attempt to make the PlayStation 5 look weaker than it is, perhaps due to Xbox fanboyism.
The PS5 TE is not a standard CU, it's generally based on a CU architecture, for instance, it has 64 SU, but it is not an RND2 CU. When Sony or MS makes a GPU, they have no way of knowing in advance which CU will be defective so as a result, they have no idea which 4 CU they will disable, it's different for each APU that gets made. So Sony needs all 40 CU to be identical in order for the active 36 CU to be identical, that's why TE can't be one of the 40 CU. If anything, the TE will probably add more power to the PS5 because it's programable, which means some developers will "hack" it and use it for other purposes.
 
Here we go with the baseless accusations again... I watched the presentation and I understand what he means. Do you?


Where does he say that they can reach sequential read speeds with random reads?


Funny... The majority that are speaking for the PS5's unique design didn't even look into the XSX apparently... Let me just copy a few things...

"DirectStorage is an all new I/O system designed specifically for gaming to unleash the full performance of the SSD and hardware decompression. It is one of the components that comprise the Xbox Velocity Architecture. Modern games perform asset streaming in the background to continuously load the next parts of the world while you play, and DirectStorage can reduce the CPU overhead for these I/O operations from multiple cores to taking just a small fraction of a single core; thereby freeing considerable CPU power for the game to spend on areas like better physics or more NPCs in a scene. "

Sound at all like PS5's solution? What about this?

" Hardware decompression is a dedicated hardware component introduced with Xbox Series X to allow games to consume as little space as possible on the SSD while eliminating all CPU overhead typically associated with run-time decompression. It reduces the software overhead of decompression when operating at full SSD performance from more than three CPU cores to zero – thereby freeing considerable CPU power for the game to spend on areas like better gameplay and improved framerates. "

Or this...?

"A component of the Xbox Velocity Architecture, SFS is a feature of the Xbox Series X hardware that allows games to load into memory, with fine granularity, only the portions of textures that the GPU needs for a scene, as it needs it. This enables far better memory utilization for textures, which is important given that every 4K texture consumes 8MB of memory. Because it avoids the wastage of loading into memory the portions of textures that are never needed, it is an effective 2x or 3x (or higher) multiplier on both amount of physical memory and SSD performance. "


There's a reason I've been saying that most of the things the PS5 can do, the XSX can do it too. The PS5 might still have a speed advantage, but it's not exclusive tech to the PS5...

"DirectStorage is the Windows API that will be used to control what Microsoft calls the Xbox Velocity Architecture. It’s Microsoft’s approach to reducing the storage capacity that an Xbox Series X game will require, promising to load the game and its assets as quickly as possible."
DirectStorage isn't a magic get out of jail free card or some form of secret sauce hardware. It's literally an API and one that'll arrive on PC sometime after Xbox Series X releases. It's a key part of the console yes, stemming from the fact that it's one of the major components of the complete API package built into the consoles' hardware. It ain't going to magically give the XSX a custom I/O block like the PS5, it's just there to allow for loading game assets faster. Take a look in this link if you're interested. It does have a have a custom decompression block sure which makes decompression in Zlib and allow for faster decompression of textures using Microsoft's own BCPack. It ain't complete hardware, it's just optimised code built into the API the console uses, nothing else.
 

DaGwaphics

Member
But the PS5 ssd is still undoubtedly in ballpark off a single channel ddr3 setup.

Not really, no. Performance, if tasked with the normal workload of traditional ram, would be less than SDR variants from the 90s. First the lack of bit level writes would kill everything, secondly the general use pattern would highlight the weaknesses of SSD not the strengths.
 

DeepEnigma

Gold Member
The PS5 TE is not a standard CU, it's generally based on a CU architecture, for instance, it has 64 SU, but it is not an RND2 CU. When Sony or MS makes a GPU, they have no way of knowing in advance which CU will be defective so as a result, they have no idea which 4 CU they will disable, it's different for each APU that gets made. So Sony needs all 40 CU to be identical in order for the active 36 CU to be identical, that's why TE can't be one of the 40 CU. If anything, the TE will probably add more power to the PS5 because it's programable, which means some developers will "hack" it and use it for other purposes.

I am in agreement, however didn't Cerny say it will contain no GPU properties like a CU? That it's more an SPU like in the Cell, which then I suppose could be used for other things besides audio if they are allowed to. SPUs were able to do adaptive tesselation as an example.
 

Goncas2

Member
As for cache scrubbing... I'll leave this here, and let people themselves draw conclusion on what the influence is on performance... And whether it really is some fancy PS5 only tech...


That Wikipedia article is about error correction in caches. PS5's cache scrubbers are very different, they avoid the need to flush (i.e. erase) the entire GPU cache whenever the SSD is read.

Very different technologies. With this technology the GPU compute units don't need to wait to get data from memory as often.
 

Shmunter

Member
Not really, no. Performance, if tasked with the normal workload of traditional ram, would be less than SDR variants from the 90s. First the lack of bit level writes would kill everything, secondly the general use pattern would highlight the weaknesses of SSD not the strengths.
Surely we can agree we are talking about reading assets at ddr3 speed, not traditional ram patterns of random reads and writes. The writes to these ssd’s would still be traditional file storage, save files or memory hibernation for game suspension etc.
 
That Switch shortage is really hurting those numbers from Nintendo. I'm surprised at how badly the X1 is doing in comparison to it's direct competitor.
Yea thats why ms started talking about next gen. Not much to lose honestly.

No wonder sony is in no rush to talk about next gen when they r doing these numbers and still have few major titles left for this year like last of us 2 and Ghost Of tsushima
 
Last edited:

Ascend

Member
Yes, dual channel is legit. But the PS5 ssd is still undoubtedly in ballpark off a single channel ddr3 setup. Of which many pc out there also did at the time, both are true at the same time.

But even at it’s single channel ddr3 speed, it surely must be acknowledged it can certainly offer plenty fast assets to feed the GPU In streaming scenarios, and a massive amount of it to boot.

Edit : I see you’ve added a bit about ssd sequential vs random access speed differences. I’ve never heard of this applying to ssd. It certainly doesn’t make sense without ssd having mechanical restrictions, but I’m no expert.
Putting all the transfer speeds aside, there is one other important aspect here. You can read from and write to RAM bit for bit. For SSDs, you can't. You can only read pages, and only write to blocks.

Surely we can agree we are talking about reading assets at ddr3 speed, not traditional ram patterns of random reads and writes. The writes to these ssd’s would still be traditional file storage, save files or memory hibernation for game suspension etc.
RAM is cleared and reloaded frequently depending on the assets that will be needed in the near future. An SSD would not be able to do this, because they are by default hampered in terms of writing.

"DirectStorage is the Windows API that will be used to control what Microsoft calls the Xbox Velocity Architecture. It’s Microsoft’s approach to reducing the storage capacity that an Xbox Series X game will require, promising to load the game and its assets as quickly as possible."
DirectStorage isn't a magic get out of jail free card or some form of secret sauce hardware. It's literally an API and one that'll arrive on PC sometime after Xbox Series X releases. It's a key part of the console yes, stemming from the fact that it's one of the major components of the complete API package built into the consoles' hardware. It ain't going to magically give the XSX a custom I/O block like the PS5, it's just there to allow for loading game assets faster. Take a look in this link if you're interested. It does have a have a custom decompression block sure which makes decompression in Zlib and allow for faster decompression of textures using Microsoft's own BCPack. It ain't complete hardware, it's just optimised code built into the API the console uses, nothing else.
If DirectStorage is only an API, how do they magically get performance of multiple Zen cores? From your own link;

Microsoft’s Goossen told Digitial Foundry that doing decompression on the 4K textures to match the speed of the SSD rate would have consumed three Zen 2 CPU cores, plus an additional two more just for the I/O overhead. With DirectStorage, Microsoft reduced that down to just a tenth of one core. All that CPU power can now be repurposed for other things.
 
Status
Not open for further replies.
Top Bottom